Preliminary communication / Communication

Application of surface science techniques in the studyof environmental photocatalysis: nitrogen-doped TiO2

Alexander Orlov, Mintcho S. Tikhov, Richard M. Lambert *

Chemistry Department, Cambridge University, Lensfield Road, Cambridge CB2 1EW, United Kingdom

Received 2 July 2004; accepted after revision 2 December 2005

Available online 09 September 2005

Abstract

In order to understand the origin of visible-light photocatalysis by nitrogen-doped titania, titania (anatase) films whosecomposition can be varied in a controlled way have been grown in situ and subsequently characterized by XPS, UPS, and ex situAFM. Two very different in situ methods for nitrogen-doping were used, N+ implantation and nitride oxidation, and the chemi-cal identity and depth distributions of the resulting chemical species were established. It is found that N+ implantation results inextensive reduction of the titania, re-oxidation with O2 establishing an essentially oxidized surface. Both methods of preparationresult in the nitrogen being located beneath the surface. These results have significant implications for our understanding for theorigin of visible light activity of nitrogen-doped catalysts. In particular, they indicate that earlier interpretations may be over-simplified or even incorrect. To cite this article: A. Orlov et al., C.R. Chimie 9 (2006).© 2005 Académie des sciences. Published by Elsevier SAS. All rights reserved.

Résumé

Afin de comprendre l’origine de la photocatalyse à lumière naturelle par l’azote enrichi de dioxyde de titane, des films de cecomposé (anatase), dont la composition peut être changée d’une manière contrôlée, ont été synthétisés in situ et plus tardcaractérisés par XPS, UPS et microscopie à force atomique ex situ. Deux méthodes pour saturer l’azote in situ très différentesont été employées, à savoir implantation de N+ et oxydation de nitrures, l’identité des espèces chimiques et les distributions enprofondeur de l’espèce résultante ont été établies. On constate que l’implantation de N+ a pour conséquence la réduction étenduedu dioxyde de titane, la réoxydation conduisant à une surface essentiellement stœchiométrique. Les deux méthodes de prépara-tion ont pour conséquence le fait que l’azote est situé sous la surface. Ces résultats ont des implications significatives pour notrecompréhension de l’origine de l’activité à la lumière visible des catalyseurs saturés en azote. En particulier, ils indiquent que desinterprétations antérieures ont pu être trop simplistes ou sont peut-être même incorrectes. Pour citer cet article : A. Orlov et al.,C.R. Chimie 9 (2006).© 2005 Académie des sciences. Published by Elsevier SAS. All rights reserved.

* Corresponding author.E-mail address: rml1@cam.ac.uk (R.M. Lambert).

C. R. Chimie 9 (2006) 794–799

http://france.elsevier.com/direct/CRAS2C/

1631-0748/$ - see front matter © 2005 Académie des sciences. Published by Elsevier SAS. All rights reserved.doi:10.1016/j.crci.2005.03.020

Keywords: Visible light; TiO2; Photocatalysis; Nitrogen-doped; XPS; UPS; AFM

Mots clés : Lumière visible ; TiO2 ; Photocatalyse ; Dopage à l’azote ; Spectroscopie de photoélectrons X ; Spectroscopie de photoélectronsUV ; Microscopie à force atomique

1. Introduction

There is a much current interest in the developmentof TiO2-based heterogeneous photocatalysts that oper-ate under visible light irradiation. Traditional ap-proaches described in literature include coupling TiO2

with organic dye sensitizers, doping with transition met-als, and reduction of TiO2 by magnetron sputtering orby hydrogen [1]. Recently published studies describ-ing TiO2 doping by nitrogen [2–8], carbon [9–12], andsulfur [9,13] indicate that this is a very promisingapproach for the production of catalysts operating atvisible wavelengths. However, the nature and mode ofaction of these materials is still the subject of debateabout the origin of such activity. On the basis of firstprinciples calculations, Asahi et al. [5] concluded thatvisible light activity is due to nitrogen in substitutionalsites of TiO2 resulting in efficient mixing with O 2pstates. However, others have argued that a combinationof titania reduction and/or formation an isolated N 2pnarrow band [2,4,7] is principally responsible for theobserved phenomena. Some studies have also reportedthat catalytic activity is a function of nitrogen-dopinglevel [7], with increased nitrogen concentration beyondan optimal level producing a detrimental effect on vis-ible light activity. Moreover, the activity was found tobe dependent on the particular procedure used for nitro-gen incorporation. Thus although oxidation of tita-nium nitride has been proposed as a method for visiblelight active producing N-doped TiO2 [3], others havefound that creation of TiN crystal phases by high tem-perature TiO2 annealing in ammonia had a detrimentaleffect on catalyst activity [7].

To clarify the underlying phenomena that result invisible light activity of N-doped TiO2, we have devel-oped a model system amenable to detailed studies underUHV conditions. Titania films of controlled and vari-able composition have been grown in situ and subse-quently characterized by XPS, UPS, and AFM. Com-pared to single crystal surfaces of titania [14], thesepolycrystalline films are likely to be much more real-istic models of practical dispersed titania catalysts interms of crystallographic structure, morphology, and

electronic properties. In this respect our results areentirely novel and should contribute to a better under-standing of photocatalytic effects involving these inter-esting materials.

2. Experimental

Experiments were performed in a VG UHV system,previously described [15] which incorporated facilitiesfor true angle resolved photoelectron spectroscopy. XPspectra were obtained with Mg Ka radiation and UPSexperiments were performed with a He I source(hm = 21.2 eV). AFM topographic images were col-lected in air at 22 ± 1°C with a Digital InstrumentDimension 3100 AFM with a Nanoscope IV controllerin tapping mode. Ultra-sharp MikroMasch (MikroM-asch Eesti OU, Estonia) silicon cantilevers (NSC15 se-ries, 125 mm long, tip radius < 10 nm, spring constant~ 40 N/m, resonant frequency ~300 kHz) were used.The images were collected at 512 × 512 pixels perimage at a scan rate of 0.5 Hz. The captured imageswere analyzed with the Nanoscope image analyzingsoftware (version 5.12r3, Veeco, CA) using first orderflattening. The 10 × 10 × 1 mm polycrystalline tita-nium sample (Advent, 99.6%) was mechanically pol-ished and cleaned in vacuum by Ar+ sputtering/annealing cycles until no impurities were detectable byXPS. A polycrystalline TiO2 thin film was then pro-duced by controlled oxidation in 6 × 10−6 mbar of oxy-gen at 500 K for 8 min followed by cooling to 300 K inthe same oxygen atmosphere.

3. Results and discussion

AFM revealed that the surface morphology of thepolycrystalline titania thin film (Fig. 1) was dominatedby featureless flat regions with a z-roughness on theorder of 1 nm. A minority component of the surfaceconsisted of nanocrystallites with z-roughnessof ≥ 3 nm.

Two different procedures were employed for gener-ating N-doped titania films. Firstly, nitridation of metal-

795A. Orlov et al. / C. R. Chimie 9 (2006) 794–799

lic Ti by exposure to N2 gas (T = 473°K, 1400 l) fol-lowed by oxidation of the resulting nitride film with O2

gas (T = 493°K, 2880 l). The second procedure involvednitrogen implantation of a titania film, prepared asdescribed above, by bombardment with N+ ions (4 keV,fluence ~ 3 × 1015 ions per cm2). It is important to rec-ognize that in addition to implanting N, XPS showedthat this process results in partial reduction of the tita-nia. Therefore, in order to produce a sample whoseproperties are relevant to those of photocatalysts work-ing in an oxygen-containing environment, it was essen-tial to re-oxidize this N-implanted titania according tooxidation procedure described above.

A comparison of results for the two procedures isillustrated in Fig. 2, which shows relevant Ti 2p and N1s XP spectra. It is apparent that nitridation by nitro-gen adsorption on metallic titanium followed by oxida-tion resulted in a somewhat lower average concentra-tion of nitrogen in the final material (as indicated by N1s intensities) compared to implantation by N+ bom-bardment. The average levels of nitrogen-doping fornitrided/oxidized and nitrogen implanted oxide filmswere 1.0% and 1.2%, respectively, as determined byXPS. These values are based on the approximation ofuniform nitrogen distribution within the sampling depthof the XPS measurement (see below). The measured N1s binding energy was 396.6 eV in both cases, with aslightly broader FWHM for the nitrided/oxidizedsample, indicating some differences in nitrogen chemi-cal environments depending on-doping procedure. A

much more pronounced difference was observed in thecorresponding Ti 2p spectra shown on Fig. 2I. The bot-tom spectrum (A) corresponds to a nitrided/oxidizedsample; spectrum (B) was obtained after implantingnitrogen into a fully oxidized TiO2 sample; spectrum(C) was recorded after re-oxidation of sample (B) andspectrum (D) was obtained after oxidation of clean tita-nium foil and corresponds to fully oxidized TiO2. Thecorresponding N 1s spectra are shown in part II of theFig. 2. Detailed analysis of the data indicated that N2

adsorption on the clean Ti surface resulted in a Ti 2pcomponent shifted by about 0.7 eV and about 20%broader as compared to clean surface. This indicatesdevelopment of a TiNx (x < 1) surface phase [16]. Afteroxidation, this component remained unshifted, indicat-ing that the TiNx phase remains largely unaffected bythe procedure (Fig. 2I, spectrum A). The only differ-ence, a small broadening towards high binding ener-gies, may be interpreted as a development of a Ti–N–Omixed phase at the interface between the nitrided andoxidized phases. The intensity of the TiNx emissionafter oxidation was attenuated by about 60%, indica-tive of an on-top oxide layer of thickness of ~ 1 nm. Incontrast, N+ implantation (Fig. 2I, spectrum B) resultedin a development of a phase of variable composition(as it might be expected) characterized by a broad Ti2p component shifted by 3.6 eV from the clean surfacevalue. As is apparent from the spectrum (Fig. 2I, spec-trum B) the oxide was strongly reduced (note the very

Fig. 1. Representative AFM image of oxidized titanium film.

Fig. 2. (I) Ti 2p XP spectra of (A) N2 adsorbed Ti sample after oxi-dation; (B) TiO2 implanted with N+; (C) TiO2 implanted with N+

after re-oxidation; (D) TiO2 sample; (II) N 1s XP spectra of (A) N2

adsorbed Ti sample after oxidation; (B) TiO2 implanted with N+; (C)TiO2 implanted with N+ after re-oxidation; (D) TiO2 sample.

796 A. Orlov et al. / C. R. Chimie 9 (2006) 794–799

low Ti4+ intensity), and the low binding energy com-ponent at 454.1 eV present in the nitrided/oxidizedsample (Fig. 2I, spectrum A) was entirely absent. Weattribute this behavior to development of a fairly thickoxygen-rich TiOxNy. Re-oxidation (Fig. 2I, spectrumC) restored the initial shape of the Ti 2p envelope leav-ing only a small residual component at ~ 456 eV due tothe TiOxNy phase located at the surface, or close to thesurface. These findings are supported by the UPS resultsdescribed below. The observed trends provide a basisfor explaining the presence/absence of photocatalyticactivity of nitrogen-modified TiO2 materials. It shouldbe noted that the position of N 1s peak in our modelsystem (396.6 eV) is similar to that observed in pow-der catalyst studies (396.0 eV) [5] indicating a resem-blance of nitrogen environments. In the two cases ourresults indicate that reports attributing visible light activ-ity of nitrogen implanted samples [5] to interstitialN-doping appear to oversimplify the phenomenon. Wehave clearly shown that nitrogen implantation of tita-nia results in a substantial reduction of the material.This suggests that visible light activity reported in somestudies [5] cannot be exclusively attributed to the effectof nitrogen-doping and that other effects such as oxy-gen vacancies creation may be significant.

Angle resolved XP spectra provide more detailedinformation about the depth distributions of nitrogenand the oxidation states of titanium, and Fig. 3 shows

the results for the nitrided/oxidized titania film. It isvery clear (Fig. 3II) that in this material the nitrogenhas migrated beneath the surface (the N 1s emission at75° photoelectron exit is about five times less intensethan that at 35° exit). Furthermore, the Ti 2p compo-nent centered at 454.1 eV and observed in 35° spec-trum (Fig. 3I) is not present in the 75° spectrum, indi-cating that this component is well below the fullyoxidized surface, in agreement with ~1 nm thicknessof the oxide layer estimated above. This observation isof direct relevance to the TiN oxidation method of cata-lyst preparation [3], as we have now shown that thisprocedure produces surfaces dominated by TiO2. There-fore, negligible activity in visible region is to beexpected. Metallic TiN is a poor photocatalyst, and highlevels of doping can lead to transformation of TiO2 toTiN decreasing visible region photocatalytic activity [7].The opposite process, oxidative transformation of TiN→ TiO2 can lead to formation of rutile structures with-out forming anatase phases, thus actually reducing pho-toreactivity for reactions in the UV region where ana-tase is the most active phase [17].

Corresponding angle resolved XPS results for there-oxidized, N+ implanted titania film are shown inFig. 4. Once again, despite a very different preparativemethod, it is evident that the nitrogen is located beneaththe surface. However, in contrast to the previous case,the shape of the Ti 2p component does not signifi-

Fig. 3. (I) Angle resolved Ti 2p XP spectra of N2 adsorbed Ti sampleafter oxidation at various detection angles: (A) detection at 35°; (B)grazing detection at 75°. (II)Angle resolved N 1s spectra of N2 adsor-bed Ti sample after oxidation at various detection angles: (A) detec-tion at 35°; (B) grazing detection at 75°.

Fig. 4. (I) Angle resolved Ti 2p XP spectra of TiO2 implanted withN+ after re-oxidation at various detection angles: (A) grazing detec-tion at 75°; (B) detection at 35°. (II) Angle resolved N 1s spectra ofTiO2 implanted with N+ after re-oxidation at various detection angles:(A) grazing detection at 75°; (B) detection at 35°.

797A. Orlov et al. / C. R. Chimie 9 (2006) 794–799

cantly change with the emission angle, indicating thatthe sample composition is rather homogenous. Weshould also note that in the fully oxidized TiO2 sample,prepared in the absence of nitrogen, the feature at~455 eV, attributed to TiOx (x < 2), is absent at thisemission angle [15], in contrast to the N+ implantedsample. Additionally, the N 1s intensity at 75° emis-sion angle is about 1/3 of that at 35° emission angle,indicating that nitrogen is close to the surface.

UP spectra of clean Ti, a TiO2 thin film, and bothtypes of nitrogen-modified TiO2 films are shown inFig. 5 while the corresponding work function and band-gap values are given in Table 1. It should be noted thatthe UPS/He-I sampling depth (0.5 nm) is significantlyless than that for Ti 2p/N 1s photoelectrons (Ek ~ 800 eVfor Mg Ka; sampling depth ~ 1.5 nm). For the pureTiO2 film, the valence band consists of features derivedfrom O 2p orbitals whereas the lower portion of emptyconduction band is composed primarily of Ti 3d orbit-als [18]. The observed band gap of 3.36 eV and themeasured work function of 5.35 eV are in excellentagreement with literature values reported for anatase

form of TiO2 [15,18] This is an important observationbecause it is generally found that anatase is a much moreeffective photocatalyst than rutile [17]. We may there-fore conclude that our experiments were carried out onthe catalytically relevant form of TiO2. The nitrided/oxidized film exhibited the UP spectra, bandgap andwork function values very similar to those of the origi-nal TiO2 surface – in very good accord with the XPSdata showing that the surface region of the film con-tains mainly Ti4+. In contrast, the N+ implanted,re-oxidized film had a smaller work function (4.66 eV),with the onset of the oxygen-derived valence featuresshifted towards higher binding energies by 0.3 eV. Italso had a much smaller bandgap (0.1 eV) as com-pared to TiO2 (3.36 eV). The dramatically decreasedbandgap is associated with the photoemission featurethat shows a maximum at ~ 0.7 eV below EF, extend-ing to 1.3 eV below EF. This feature may be assignedto a partial occupation of the Ti 3d band (3dx, x > 0).Similar observations have been reported for bulk TiOx

(x < 2) where a decrease in work function, partial occu-pation of the Ti 3d band extending to 1.5 eV below EF,and a shift of the O 2p band to higher binding energies[19,20] were found. This partial filling of the Ti 3d bandcannot be unequivocally assigned to the presence ofoxygen vacancies as similar emission is observed inthe case of non-stoichiometric TiN0.5 [16].

To rationalize our observations, two distinct sce-narios resulting from two different methods of nitro-gen incorporation into TiO2 may be proposed. Thedevelopment of a titanium nitride thin film for N2

adsorbed/oxidized titanium sample resulted in substan-tial changes in XPS features attributed to non-stoichiometric titanium nitride. UPS, a more surfacesensitive technique than XPS, confirmed developmentof TiO2 overlayer after subsequent oxidation. UPSresults also indicated that this overlayer has the char-acter of fully oxidized TiO2 with titanium nitride spe-cies having negligible contribution to the spectra. Onthe other hand, in accord with XPS results, N+ im-planted samples developed titanium oxynitride and/orreduced titania species much closer to the surface, trans-lating into significant changes in bandgap and workfunctions values as determined by UPS. It is apparentthat the effects of nitrogen incorporation stronglydepend on preparation conditions and are a complexfunction of titanium oxidation state and nitrogen spe-cies distribution throughout the sample depth. These

Fig. 5. UPS spectra of Ti, TiO2, and nitrogen-modified TiO2 sam-ples.

Table 1Bangap and work function values of the samples as determined byUPS

Sample Bandgap (eV) Workfunction(eV)

TiO2 3.36 5.35TiO2 N+ sputtered and oxidized 0.1 4.66Ti N2 adsorbed and oxidized 3.29 5.35Ti 0 3.59

798 A. Orlov et al. / C. R. Chimie 9 (2006) 794–799

parameters can strongly affect the photocatalytic activ-ity of the resulting materials in the visible region. Thedevelopment of optimized and reproducible methodsof N incorporation is an essential requirement for theproduction of reliable, stable and visible light activephotocatalyst. Possible future strategies include co-dosing O2 and N2 using reactive gas pulsing tech-niques [21], for example.

4. Conclusions

1. Polycrystalline, stoichiometric anatase thin filmsmay be grown by controlled oxidation of Ti metal undervacuum conditions and doped with nitrogen by N+ bom-bardment. The resulting material is substantially re-duced, re-oxidation resulting in the incorporated nitro-gen being located close to the surface.

2. An alternative chemical route, nitridation of themetal followed by oxidation, results in an anatase filmwith a similar degree of N-doping. In this case, how-ever, the nitrogen is located well below the surface.

3. The results imply that earlier interpretations ofthe origin of visible light activity of nitrogen-doped pho-tocatalysts may be oversimplified or even incorrect. Inparticular, the role of oxygen vacancies, inevitably intro-duced by N+ implantation, should be given more sig-nificance.

Acknowledgements

This work was supported under Research GrantGR/R03082/01 awarded by the UK Engineering andPhysical Sciences Research Council. A.O. thanks theEnvironmental Research and Education Foundation forthe award of an International Scholarship, and Schlum-berger Cambridge Research for additional financial sup-port.

References

[1] M. Anpo, M. Takeuchi, J. Catal. 216 (2003) 505.

[2] H. Irie, Y. Watanabe, K. Hashimoto, J. Phys. Chem. B 107(2003) 5483.

[3] T. Morikawa, R. Asahi, T. Ohwaki, K. Aoki, Y. Taga, Jpn J.Appl. Phys. Part. 2 – Lett. 40 (2001) L561.

[4] T. Ihara, M. Miyoshi, Y. Iriyama, O. Matsumoto, S. Sugihara,Appl. Catal. B-Environ. 42 (2003) 403.

[5] R. Asahi, T. Morikawa, T. Ohwaki, K. Aoki, Y. Taga, Science293 (2001) 269.

[6] S. Sakthivel, H. Kisch, Chem. Phys. Chem. 4 (2003) 487.

[7] M. Miyauchi, A. Ikezawa, H. Tobimatsu, H. Irie, K. Hash-imoto, Phys. Chem. Chem. Phys. 6 (2004) 865.

[8] T. Sano, N. Negishi, K. Koike, K. Takeuchi, S. Matsuzawa, J.Mater. Chem. 14 (2004) 380.

[9] H. Irie, Y. Watanabe, K. Hashimoto, Chem. Lett. 32 (2003)772.

[10] S.U.M. Khan, M. Al-Shahry, W.B. Ingler,, Science 297 (2002)2243.

[11] S. Sakthivel, H. Kisch, Angew. Chem. Int. Ed Engl. 42 (2003)4908.

[12] Z.G. Zou, J.H.Ye, K. Sayama, H. Arakawa, Nature 414 (2001)625.

[13] T. Ohno, M. Akiyoshi, T. Umebayashi, K. Asai, T. Mitsui,M. Matsumura, Appl. Catal. A-Gen. 265 (2004) 115.

[14] O. Diwald, T.L. Thompson, E.G. Goralski, S.D. Walck,J.T. Yates, J. Phys. Chem. B 108 (2004) 52.

[15] E.C.H. Sykes, M.S. Tikhov, R.M. Lambert, J. Phys. Chem. B106 (2002) 7290.

[16] M. Guemmaz, G. Moraitis, A. Mosser, M.A. Khan, J.C. Par-lebas, J. Phys. Condens Mater 9 (1997) 8453.

[17] M.R. Hoffmann, S.T. Martin, W.Y. Choi, D.W. Bahnemann,Chem. Rev. 95 (1995) 69.

[18] A.L. Linsebigler, G.Q. Lu, J.T. Yates,, Chem. Rev. 95 (1995)735.

[19] V.E. Henrich, H.J. Zeiger, T.B. Reed, Phys. Rev. B 17 (1978)4121.

[20] Q. Guo, W.S. Oh, D.W. Goodman, Surf. Sci. 437 (1999) 49.

[21] N. Martin, R. Sanjines, J. Takadoum, F. Levy, Surf. Coat.Technol. 142 (2001) 615.

799A. Orlov et al. / C. R. Chimie 9 (2006) 794–799